In Hunting the elusive search strategy Jon Udell proposes to study the methods that people use to effectively search. He proposes capturing individual cases of searching for something and then analyzing those cases.
The point is to help people learn to search.
That’s an interesting approach, and it would benefit from having some good tools to capture the steps of a particular search in a way that the process could be simple annotated. I assume that Google is constantly mining their users search histories to improve their results. Of course, they don’t have any direct access to the users thoughts (or do they?, bwahahaha) but many cases are probably pretty obvious, especially if you can see where the user eventually ended up.
On the other hand, I already have conscious mental model that I use when searching. Maybe others would benefit by it. It has the advantage of being a concrete physical task instead of an abstract logical or mathematical concept. This kind of model is easier to grasp in your mind because it is actually possible to grasp in your hands. Having held the objects in your hands and manipulated them demystifies the process. There is an enormous area of your brain devoted to your hands. Let’s use it.
The mental model comes from an old information retrieval technology called edge-notched cards. When I was a child you could purchase special index cards for research note taking and filing. The cards were about 5 x 8 inches and had rows of holes along three edges. There were two or three holes deep and probably 40-50 columns of holes along the top and 20-30 along the sides. The center was lined like a conventional index card. A corner was cut off so that if you had a deck of cards, you could be sure that all the cards were oriented the same way.
The idea was that you would take notes on subject and then “file” the cards by using a special cutter to notch the holes into slots. You would search by inserting a rod into the holes of the card that you had decided would be associated with the search term and lifting the deck. Those cards that match your search term were notched and would fall from the deck. If you wanted to narrow your search further, you could repeat that search on your smaller deck. If you wanted to widen your search, you could go back to the main deck and pull out more search terms.
The power of this approach was that the “terms” could be anything. You could have a term be “Rasputin” or your term could be “Stories of the debauchery of Catherine the Great,” or it could be 1917. You could have as many index terms as you had holes in the card. The downside is that you had to have a pretty good idea what your terms ought to be when you started and that you could run out of holes. I don’t know how these problems played out in practice.
There is something about this concrete metaphor that is very useful and accessible. I know about databases, set operations, boolean logic, tagging and the mathematics of searching, but this concrete image of cards in a deck is what comes back to me when I am searching.
It also gives rise to a mental model of what the search landscape looks like. There are certain areas or the landscape that are easy to find. If I search for my name, or this domain name, there aren’t that many documents returned. Searching is easy when you have terms that return a small number of documents.
As I have blogged before about searching, it is normally pretty easy to find a very specific phrase, such as an error message. Just quote the non-varying part of the message in a search, and magically pull the right result out of the web. This is becoming more difficult as the search landscape is polluted by application trace files and logs. (As an aside, my experience is that it is becoming more and more common to do a very specific search and to find a log file where the same error is seen, but without any assistance. Google’s decision to index log files increases the noise of the search space.)
The search for the name “Edge-notched cards” was hard. I was sure that the cards had a name. My experience with them was seeing them sold in the pages of a Heath Kit catalog, which must have called them something. There are many pages that refer to “index cards” and there is no shortage of pages that refer to both “index cards” and “hole punch”. The landscape is busy in these areas. Apparently both office suppliers and teachers are prolific producers of web pages that use both of these terms.
By searching for “index card holes filing” (no quotes) I did manage to find this page where Matt Neuberg reviews a product called SlipBox. Apparently Matt is using the same mental model for search as I am, and he has the advantage of having used the cards for a significant project, but if knows what they are called, he isn’t telling. Adding “search” to the terms, moves Matt up the page, that’s a good sign, and another rule for Jon’s list, but doesn’t add anything useful.
A Wikipedia article, Index Card from searching ‘”index cards” hole information search’ says that these are called “Needle cards” and asserts that they were the precursor of Hollerith (“punched”) cards. The article also asserts that needle cards were a 20th century innovation. As I know independently that Hollerith cards were used in the 1890 census, I see that the article isn’t entirely consistent. I’m also dubious of the claim that needle cards were precursors of punched cards in any case. The cards are the same, but the technology is not. They are more like two forks of a tree instead of lineal descendants.
A Google search for “Needle cards” refers most frequently to the cards that needles are sold on and apparently are collectable. (No help here.)
Going back to the Wikipedia entry for Punch Cards there was a reference to the Wikipedia entry Edge-notched card, and here we have a winner. These were invented in 1896 (so I was right about them not being a precursor of punch cards). Here is the orignial source for the above picture.
A variety of names, generic and trademarked, have been used for marginal punched-card systems: edge-notched cards, slotted cards, E-Z Sort, Zatocards, McBee, McBee Keysort, Flexisort, Velom, Rocket, and many others